Model summary and visualization in TensorFlow - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of model.summary() for a simple CNN

Consider this TensorFlow Keras model:

model = tf.keras.Sequential([
  tf.keras.layers.Conv2D(16, 3, activation='relu', input_shape=(28,28,1)),
  tf.keras.layers.MaxPooling2D(2),
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(10, activation='softmax')
])
model.summary()

What is the total number of trainable parameters shown in the summary?

TensorFlow

import tensorflow as tf
model = tf.keras.Sequential([
  tf.keras.layers.Conv2D(16, 3, activation='relu', input_shape=(28,28,1)),
  tf.keras.layers.MaxPooling2D(2),
  tf.keras.layers.Flatten(),
  tf.keras.layers.Dense(10, activation='softmax')
])
model.summary()

A 27210

B 14890

C 15000

D 14796

💡 Hint

Calculate parameters for Conv2D: (kernel_height * kernel_width * input_channels + bias) * filters, then add Dense layer params.

Conv2D layer has (3*3*1 + 1) * 16 = 160 parameters. Flatten layer has no parameters. Dense layer has (13*13*16)*10 + 10 = 27050 parameters. MaxPooling reduces spatial size from 26x26 to 13x13. Total parameters = 160 + 27050 = 27210.

🧠 Conceptual

intermediate

Purpose of model.summary() in TensorFlow

What is the main purpose of calling model.summary() in TensorFlow Keras?

A To visualize the training loss and accuracy graphs

B To print a detailed layer-by-layer description including output shapes and parameter counts

C To save the model architecture to a file

D To train the model on the dataset

💡 Hint

Think about what information you get immediately after building a model.

model.summary() prints a table showing each layer's name, output shape, and number of parameters. It helps understand model complexity before training.

❓ Model Choice

advanced

Choosing the correct code to visualize a model architecture

Which code snippet correctly generates a plot image file named 'model_plot.png' that visualizes the TensorFlow Keras model architecture?

TensorFlow

import tensorflow as tf
from tensorflow.keras.utils import plot_model

model = tf.keras.Sequential([
  tf.keras.layers.Dense(32, input_shape=(10,), activation='relu'),
  tf.keras.layers.Dense(1, activation='sigmoid')
])

A plot_model(model, to_file='model_plot.png', show_shapes=True)

B model.save('model_plot.png')

C model.plot('model_plot.png')

D plot(model, filename='model_plot.png')

💡 Hint

Look for the utility function designed for plotting models in Keras.

The function plot_model from tensorflow.keras.utils is used to create a visualization image file of the model. Other options are invalid or do not exist.

❓ Metrics

advanced

Interpreting model.summary() output for parameter counts

In the model summary output, what does the 'Param #' column represent for each layer?

A The number of input features to that layer

B Only the number of trainable parameters in that layer

C The total number of trainable and non-trainable parameters in that layer

D The number of output neurons in that layer

💡 Hint

Think about what parameters include in neural networks.

The 'Param #' column shows the total parameters including weights and biases, both trainable and non-trainable if any, for each layer.

🔧 Debug

expert

Diagnosing why plot_model does not generate an image

You run this code:

from tensorflow.keras.utils import plot_model
plot_model(model, to_file='model.png', show_shapes=True)

But no image file is created and no error is shown. What is the most likely cause?

A The model variable is not defined

B The to_file parameter must be an absolute path

C show_shapes=True causes the function to fail silently

D Graphviz or pydot package is not installed or not configured properly

💡 Hint

Check if external dependencies for plotting are installed.

plot_model requires Graphviz and pydot to generate images. If missing, no file is created but no error is raised.